This markdown summarises the output from pyfMRIqc (https://github.com/DrMichaelLindner/pyfMRIqc), a tool develeoped by Michael Lindner & Brendan Williams. This toolbox takes a 4D timeseries and computes different QC parameters.

I used the raw nifti files (as they came out of the scanner), but concatenated them so that the time series only includes volumes covering the task. This was necessary because we had to stop the scanner manually at the end of the task.

All images share the following scan paramters:

Additionally, I used a mask for the SNR calculations. The mask was created was each timeseries specifially using AFNI’s 3dAutomask. 25 was used as SNR threshold, i.e. percentage of voxel with the lowest values (outside the mask) for SNR calculation (scalar value). SNR is calculated as mean voxel intensity over time (within the brain mask) divided by standard deviation of the mean of noise. Mean noise is defined as average of the 25% (i.e. SNR threshold) of voxel having the lowest mean intensity (outside the brain mask). The SNR provides an estimate of the reliability (~ reproducibility) of fMRI data and serves as a general goodness measure.

Motion parameters have been calculated using AFNI’s 3dvolreg. This file containes the absolute motion relative to the first volume. Relative motion was then calculated with pyfMRIqc.

In total, there are the following output parameters:

Below, each of these parameters will be defined. Additionally, for each of these parameters, four graphs have been created to explore them in the sample:

  1. Histogram for all scans
  2. Histogram divided by group (control vs experimental) and task (magictricks vs rest)
  3. Boxplot for each acquisition (magictricks_run-1 vs. magictricks_run-2 vs. magictricks_run-3 vs. rest_run-1 vs. rest_run-2)
    In the boxplot, outliers are determined as: value < quantile(value, 0.25) - 1.5 * IQR(value) | value > quantile(value, 0.75) + 1.5 * IQR(value). If a value was an outlier, it was labeled using subject ID. The labels are printed in black if this subject belongs to the control group and in grey if the subject belongs to the experimental group.
  4. Spaghettiplot highlighting the timecourse of each of these parameters for each subject
    In the spaghetti plot, a subject was labeled if any of the values over time were larger than mean(value) + 1SD(value). Please note that the line for subject 16 makes a little dip for magictricks_run-2. This is because the scanner had to be stopped during the second block, so there are two acqusitions (both covering approximately half of the block).

Summary Mean

Definition: Mean voxel (all voxels) intensity over time using all voxels. The tool also provides a 2D png image (axial slices) as output for the mean voxel intensity. It can be used to see if there is any general signal loss.

Interpretation: Mean voxel intensity across all voxels is fairly normally distributed across the sample as well as within each task and group. The mean and SD are also comparable. The outliers sub-control003 and sub-experimental014 have been screened in more detail, but the mean image does not suggest any signal loss for either of them. sub-experimental014 has a large head which could have contributed to high mean values. The spaghetti plot suggests that the mean voxel intensity across time is similar across scans, so that there is no evidence scanner issues during the course of an experiment. This is also reflected by a low mean and SD of the subjectwise SD, i.e. the SD of the values for each subject.

Mean of subject-wise mean: 116.4036
SD of subject-wise mean: 21.5016
Mean of subject-wise SD: 1.9166
SD of subject-wise SD: 1.5045

Summary Mean_(mask)

Definition: Mean voxel intensity over time using voxels inside the brain mask only.

Interpretation: Results observed in the mean voxel intensity across all voxels are mirrored in the mean voxel intensity within the brain mask. When only looking at brain voxels, the mean intensity is substantially higher as expected and dis not affected by task or group. Again, the values are fairly constant across the duration of each experiment. This is also reflected by a low mean and SD of the subjectwise SD, i.e. the SD of the values for each subject.

Mean of subject-wise mean: 456.6868
SD of subject-wise mean: 48.5843
Mean of subject-wise SD: 8.0338
SD of subject-wise SD: 5.5281

Summary SD

Definition: Standard deviation of voxel intensity over time using all voxels. The tool additionally outputs a 2D png image (axial slices). By default, the grey matter, brain stem and blood vessels have the highest variability of BOLD signal. Therefore, these tissue types have higher values and are brighter in this image. Sudden or unexpected signal changes such as artefacts, signal changes, motion etc. in one or a few volumes will increase the variance and therefore will be detectable in these images.

Interpretation: The SD of voxel intensity across time for all voxels follows a normal distribution and does not differ across tasks and groups. The values are constant across scans for each subject, reflected by low mean and SD of the subjectwise SD. Again, sub-control003 and sub-experimental014 are outliers, but examining their variance images does not flag up potential artefacts.

Mean of subject-wise mean: 223.0311
SD of subject-wise mean: 30.4324
Mean of subject-wise SD: 4.0429
SD of subject-wise SD: 2.8891

Summary SD_(mask)

Definition: Standard deviation of voxel intensity over time using voxels inside the brain mask only.

Interpretation: The SD of voxel intensity for voxels within the brain mask reflects the pattern observed previously. There are no differences depending on group and task and the values are constant within each subject.

Mean of subject-wise mean: 200.7709
SD of subject-wise mean: 29.6709
Mean of subject-wise SD: 4.9906
SD of subject-wise SD: 3.7669

Summary SNR_voxel_MEAN

Definition: Mean values of voxel for SNR (I assume this is the mean voxel intensity over time of the voxels outside the brain mask and below SNR threshold, i.e. noise).

Interpretation: The mean of the voxels used for SNR calculation do not differ depending on group or task. sub-experimental018 values are outliers. After looking at the SNR mask (i.e. a mask showing with values have been used for the SNR calculation), this is due to the fact that a high number of voxels was classified as noise, some including skull and other non-brain tissues. Again, the values are constant across the scans for each subject.

Mean of subject-wise mean: 12.6095
SD of subject-wise mean: 0.9728
Mean of subject-wise SD: 0.1467
SD of subject-wise SD: 0.123

Summary SNR_voxel_STD

Definition: Standard deviation of voxel for SNR (I assume this is the tandard deviation of voxel intensity over time of the voxels outside the brain mask and below SNR threshold, i.e. noise).

Interpretation: The SD of the voxels for SNR calculation does not differ depending on group or task. Again, sub-experimental018 is an outlier, but this seems to be due to the extend of the mask. The values are again constant across scans for each subject.

Mean of subject-wise mean: 3.3593
SD of subject-wise mean: 0.3129
Mean of subject-wise SD: 0.0431
SD of subject-wise SD: 0.0367

Summary Min_Slice_SNR

Definition: Lowest slice SNR value (within the brain mask). The slice SNRs measures the time course SNR averaged across each slice. The higher the SNR, the smaller the relative fluctuations and more stable is the signal over repeated measurements.

Interpretation: The lowest slice SNR value is slightly lower in the control group compared to the experimental group, but is comparable for both tasks. Some subjects have fairly high values as their lowest SNR slice (i.e. sub-experimental034, sub-experimental028, sub-control015, sub-control041). The min slice SNR shows some variability across subjects, but this could potentially be attributed to the fact that participants might have moved in the breaks between the scans, so that a slice does not necessarily cover the same brain regions across scans.

Mean of subject-wise mean: 346.31
SD of subject-wise mean: 102.2487
Mean of subject-wise SD: 28.5949
SD of subject-wise SD: 23.7212

Summary Max_Slice_SNR

Definition: Highest slice SNR value (within the brain mask). The slice SNRs measures the time course SNR averaged across each slice. The higher the SNR, the smaller the relative fluctuations and more stable is the signal over repeated measurements.

Interpretation: The hightest slice SNR is slightly higher in the control group compared to the experimental group, but is comparable for both tasks. sub-experimental034 is an outlier. The max slice SNR values are more constant across scans for each subject than the min slice SNR values.

Mean of subject-wise mean: 833.7394
SD of subject-wise mean: 210.0937
Mean of subject-wise SD: 23.7801
SD of subject-wise SD: 11.4415

Summary Mean_voxel_SNR

Definition: Mean voxel SNR is the average over all the slices together

Interpretation: The SNR averaged across slices follows a normal distribution and is similar across tasks and groups though the experimental group has higher SDs. Again, sub-experimental034 is an outlier. The values are constant across scans for each subject.

Mean of subject-wise mean: 568.0828
SD of subject-wise mean: 138.9635
Mean of subject-wise SD: 17.3205
SD of subject-wise SD: 16.3794

Summary Mean_absolute_Movement

Definition: Absolute movement captures the difference in relation to the base volume. The values have been transformed into absolute values before calculating the mean. The mean has been calculated over all volumens and across all 6 motion parameters.

Interpretation: The mean absolute movement is constant across tasks and groups. Some subjects show a high value, either in the first task run or first rest run. There is considerable heterogeneity across scans showing that the motion was not constant across the duration of the experiment.

Mean of subject-wise mean: 0.2843
SD of subject-wise mean: 0.0962
Mean of subject-wise SD: 0.1064
SD of subject-wise SD: 0.0593

Summary Max_absolute_Movement

Definition: Maximum absolute movement (in absolute values) across all volumens and motion parameters.

Interpretation: The max absolute movement values are unfortunately quite high, some even exceeding voxel size. This affects both tasks and both groups. Again, there is a lot of variation across the experiment for each subject.

Mean of subject-wise mean: 1.2466
SD of subject-wise mean: 0.5159
Mean of subject-wise SD: 0.4795
SD of subject-wise SD: 0.269

Summary Max_relative_Movement

Definition: Relative movement captures the difference to the previous volume (rather than to the base volume). Maximum value of relative movement (in absolute values) across all volumens and motion parameters.

Interpretation: The max relative movement values are luckily smaller and do not exceed voxelsize. Outliers can be found across groups and tasks. As previously, movement changes throughout the course of the experiment.

Mean of subject-wise mean: 0.5597
SD of subject-wise mean: 0.363
Mean of subject-wise SD: 0.2284
SD of subject-wise SD: 0.1809

Summary Relative_movements_(>0.1mm)

Definition: Absolute values of relative movement are compared against threshold of 0.1mm. Output captures number of movements, rather than number of slices affected.

Interpretation: When interpreting these numbers, it needs to be considered that the reflect the number of movements in any of the 6 slices rather than the number of volumes affected. Some subjects have a high number of movements > 0.1mm, but again, both groups and tasks are affected. Again, there are changes throughout the duration of the experimentn and it seems that people were moving more as the experiment proceeded.

Mean of subject-wise mean: 83.7907
SD of subject-wise mean: 78.3669
Mean of subject-wise SD: 41.6462
SD of subject-wise SD: 42.1584

Summary Relative_movements_(>0.5mm)

Definition: Absolute values of relative movement are compared against threshold of 0.5mm. Output captures number of movements, rather than number of slices affected.

Interpretation: When looking at relative movements > 0.5mm, luckily the average number drops and again, the number reflects movements in any of the 6 parameters rather than slices affected. Both tasks and groups are affected and values change throughout the experiment.

Mean of subject-wise mean: 3.59
SD of subject-wise mean: 8.2476
Mean of subject-wise SD: 2.8594
SD of subject-wise SD: 5.5912